Mathematical Reasoning on AIME25 (Accuracy, Efficiency k)
46.7Accuracy (%)Qwen3-4B-Inst-2507
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen3-4B-Inst-2507Chat Template=Off, Decoding=Baseline, Step=0, k=12026.02 | 46.7 | 1 | |
| Qwen3-4B-Inst-2507Chat Template=Off, Decoding=ConfAdapt, tau=0.9952026.02 | 26.7 | 2.3 | |
| Qwen3-4B-Inst-2507Chat Template=Off, Decoding=Static, k=12026.02 | 23.3 | 1 | |
| Qwen3-4B-Inst-2507Chat Template=Off, Decoding=ConfAdapt, tau=0.952026.02 | 23.3 | 2.8 | |
| L3.1-8B-MagpieChat Template=On, Decoding=Baseline, Step=0, k=12026.02 | 0 | 1 | |
| L3.1-8B-MagpieChat Template=On, Decoding=Static, k=12026.02 | 0 | 1 | |
| L3.1-8B-MagpieChat Template=On, Decoding=ConfAdapt, tau=0.92026.02 | 0 | 5.5 |